How can current quantum computers solve complex industrial optimization problems that far exceed the qubit capacity of a single processor?This use case demonstrates a scalable hybrid quantum-classical pipeline for thousand-variable constrained optimization. Applied to Minimum Birkhoff Decomposition (MBD) — a challenge in network scheduling, optical switching, and energy trading — our approach combines spectral graph partitioning, shallow QAOA sampling, classical feasibility repair, and Extended Fully-Corrective Frank-Wolfe (E-FCFW) optimization.We validated the framework on the 1,354-bus PEGASE transmission grid benchmark, a 1,710-variable problem requiring 1,710 qubits without partitioning. Using spatial circuit packing on 156-qubit IBM Heron processors, independent subproblems run concurrently, achieving a 3x reduction in QPU execution time. Hardware results prove that repaired QAOA samples deliver lower residual error than simulated annealing and random sampling baselines.This work provides a proven blueprint for deploying quantum sampling on large-scale industrial problems today. We invite network operators, energy companies, logistics providers, and quantum researchers interested in scaling hybrid algorithms for real-world applications to reach out and collaborate!